THE LONGITUDINAL PROBLEM
One of the important data needs identified by the panel is to enable researchers to trace the employment of the nation’s science and engineering workforce over time. The NCSES human resources data program has several “intake” surveys that capture essential data on these individuals, including their institutions, programs of study, demographics, motivations, and attitudes, among other domains. At intake, a great deal is known about these people, and NCSES regularly collects data on new cohorts entering the science and engineering workforce. Thus the data system is a powerful resource for tracking flows into the science and engineering workforce (with the exception of immigrants educated abroad who enter the United States with similar qualifications). The Survey of Earned Doctorates (SED) and the National Survey of College Graduates (NSCG) collect data from a significant number of these highly educated people. The SED then feeds into the Survey of Doctorate Recipients (SDR), which has a longitudinal component. Because of the size of the sample, however, the SDR reduces the size of the panel component to maintain coverage breadth for its first wave.
The excellent data on flows into the science and engineering workforce constitute a good starting point. However, the careers of scientists and engineers change and end, and individuals may emigrate either permanently or temporarily. Decades ago, employment arrangements for workers in science and engineering fields were very stable—as was the case for a great many white-collar workers. The average number of career jobs held by persons
in the latter years of the baby boom (1957–1964) was 11.7 from age 18 to 48 (Angeles, 2016). Nearly one-half of these jobs were held from ages 18 to 24. Among newer entrants to the science and engineering workforce in recent years, similar job churning is likely, especially shortly after graduation, when workers are trying to place themselves in employment arrangements they find satisfactory, or at least as good as they can get.
UNEMPLOYMENT INSURANCE EARNINGS DATA AS A SUPPLEMENT TO THE CORE SCIENTISTS AND ENGINEERS STATISTICAL DATA SYSTEM (SESTAT) EFFORT
All states collect earnings data for covered employment1 for their unemployment insurance programs. This system collects quarterly earnings, the employer identification number (EIN), and the number of weeks employed for covered workers. With the EIN, one can recover the North American Industry Classification System number, which gives the type of industry in which the worker is employed. These data can be used to compute the person’s wage and salary income, mobility between employers, and location of the employer.2 With the EIN one can also recover the employer’s wage bill and the number of persons it employs. There are no occupational data. The unemployment insurance earnings data are highly comparable across states (the Census Bureau has done the work of compiling all these state data into a single structure) and include data on all jobs held, regardless of state or locality, that are covered by the unemployment insurance system. The data do not include self-employment or other income, but in the grand scheme of work for pay, the coverage is excellent, and the sources of income that are missing are missing consistently through time.
The Census Bureau has agreements with all states allowing it to use the unemployment insurance earnings data to match to census surveys. Matching the unemployment insurance earnings data to other databases would require either a new batch of memoranda of understanding with the various states or new legislation. However, these challenges are reduced for the NSCG because the data are collected by the Census Bureau, and the legal authority to match is in place. This would mean, however, that the newly matched data would be covered by Title 13. The large size of the NSCG makes the use of administrative data matching compelling if only on cost grounds.
With this matching, one could track employment, earnings, and location for the science and engineering workforce survey respondents
1 Uncovered employment, except for the self-employed, will be infrequent for SESTAT respondents.
2 Although not necessarily the location of the worker.
surveyed by the Census Bureau. This would provide reliable employment and earnings data for every respondent surveyed by the Census Bureau, and at a per-case cost orders of magnitude less than that of the current survey effort. If, at intake, respondents provided their cell phone number, then whenever a respondent changed employers, one could ask a few questions that would illuminate the job transition, such as what the survey respondent’s new occupation is and why the person left his or her previous job. Just a handful of questions would provide timely information about employment transitions. Respondents might be contacted with a few questions once every 5 years or so, with administrative data matching enabling a more passive tracking of their careers. This inexpensive tracking strategy would produce a longitudinal file that would enable analysis of career dynamics for a broader range of respondents than is currently possible. The unemployment insurance earnings data are quarterly, so the employment and earnings history information obtainable from administrative data is more detailed in terms of both quality and frequency relative to what can be attempted in the SDR.3
This administrative effort would not replace the current intake questionnaires, but supplement them. Even absent any survey effort directed at past entrants in the NCSES data system of scientist and engineers, matching in the unemployment insurance earnings data would provide significant information on the evolution of science and engineering careers as well as degree holders in other disciplines, and would do so at very modest cost. A “skinny” survey effort that asked a few trenchant questions, such as the individual’s occupation and whether his or her responsibilities are more managerial, entrepreneurial, or research and development in nature would get to the heart of what many want to know about the nation’s science and engineering human capital infrastructure. Making such abbreviated contact every 5 years or when the subject changed employers would reveal a great deal about what the nation’s stock of science and engineering workers are doing relative to research and development.
What would be lost would be the far lengthier reinterviews for prior survey intake cases. However, these reinterviews do not dominate unemployment insurance earnings data when it comes to understanding job transitions, employment, earnings, and industry. This modification would change altogether the way employment dynamics in the survey datasets are viewed
3 A data source that provides monthly rather than quarterly data is the National Directory of New Hires, a database that provides information on people taking new jobs. See https://www.acf.hhs.gov/css/resource/overview-of-national-directory-of-new-hires [January 2018].
and support examination of these dynamics across a wider variety of demographics and fields of study and a greater number of persons tracked relative to what could be afforded with the current strategy.